I still have issues with the problem that you can't tune the rate of the stream and the breadth of your catchment independently. Or at least not very effectively. But this post was a big help in crystallizing how to think about the problem... - Cameron Neylon

Cameron, agreed. Those dials need to be there for things to be really "just in time", since that "just" is not the same for everyone, but I do think one needs to think about it along the lines described - Deepak Singh from IM

Tagging + learning algorithm-based filters would be ideal. Your connections make a decent stand in for that; humans, after all, have pretty good learning. The problem is, they filter what they want, not what YOU want. - Chris Lasher

but that means you don't trust your network doesn't it? - Deepak Singh from IM

Also need to ask the question of whether you can trust your network, whether it is possible to build the right network for you at the moment. I wouldn't trust this community to bring me all the material I need on data analysis approaches for small angle scattering for instance. But something might flash past that is relevant that I wouldn't have picked up by some other means (e.g. Pawel's bookmarks from the other day I might easily have missed but they are relevant and useful) - Cameron Neylon

Also consider that popular items on networks can become lowest common denominator situations, not necessarily high information. Pertinent entries may not be "liked" by other people on my network because each of us shares limited overlaps in interest--gaps form. There's also stochasticity. Entries posted at 2 AM EST instead of at 2 PM, whether a "hub" member in your network "liked" or commented on the entries, ... It would be interesting to see if they make a difference. Anecdotally, it seems they do. - Chris Lasher

That's right, Chris. The way I always explain it is like this: Digg is useless because it's totally a slave to the majority or users on the site. Friendfeed doesn't suffer the same problem, because the rankings or tailored to your particular set of contacts. I doesn't matter if 1000000 people are talking about celebrity nonsense, because people you follow most likely won't be. - Mr. Gunn

Of course, there is an effect of timing and networks, but the overriding concept of a content-based high-throughput network is that you're finding content based on it's first or second-degree importance to you, instead of having to use someone's personality as a proxy. It doesn't solve the self-reinforcing popularity problem, but it's the best shot at mitigating it we've found so far, IMO. - Mr. Gunn

People who help me find interesting stuff, Richard Akerman, John Dupuis, Stephen Francoeur, Lambert Heller, Deborah Fitchett and more - aarontay

I add that while my network doesn't help me filter information for what interests me most, it does help me find information I would not have discovered on my own (in a timely manner). - Chris Lasher

A crucial part of trusting your network is to understand its aggregate selection bias. I trust my network for technology items, but I don't expect items of aesthetic interest to bubble. If I wanted my network to reflect my entire set of interests, I would need to tune my network to achieve a more holistic trust. - Benson Miller

Benson, that's why it's useful to have different networks in different places. My Flickr contacts are vastly different from my FF contacts. - Mr. Gunn